LLaMA-8B

Meta's 8-billion-parameter LLM (Llama 3.1): a resource-efficient, state-of-the-art model optimized for fast, multilingual dialogue and 128K context processing.

Llama 3.1 8B is Meta's smallest, most deployable model in the Llama 3.1 series (8 billion parameters). It utilizes an optimized transformer architecture with Grouped-Query Attention (GQA) for high inference efficiency, making it ideal for budget-constrained or edge applications. The model is instruction-tuned for multilingual dialogue, excels in reasoning and code generation, and features a substantial 128K token context window. Pretrained on approximately 15 trillion tokens, it delivers competitive performance against larger models in its class while maintaining superior resource efficiency.

https://llama.meta.com/llama-downloads